CN107247938A - A kind of method of high-resolution remote sensing image City Building function classification - Google Patents

A kind of method of high-resolution remote sensing image City Building function classification Download PDF

Info

Publication number
CN107247938A
CN107247938A CN201710425766.8A CN201710425766A CN107247938A CN 107247938 A CN107247938 A CN 107247938A CN 201710425766 A CN201710425766 A CN 201710425766A CN 107247938 A CN107247938 A CN 107247938A
Authority
CN
China
Prior art keywords
building
remote sensing
land
cuclear density
function
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201710425766.8A
Other languages
Chinese (zh)
Other versions
CN107247938B (en
Inventor
刘亚岚
曲畅
任玉环
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Institute of Remote Sensing and Digital Earth of CAS
Original Assignee
Institute of Remote Sensing and Digital Earth of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Institute of Remote Sensing and Digital Earth of CAS filed Critical Institute of Remote Sensing and Digital Earth of CAS
Priority to CN201710425766.8A priority Critical patent/CN107247938B/en
Publication of CN107247938A publication Critical patent/CN107247938A/en
Application granted granted Critical
Publication of CN107247938B publication Critical patent/CN107247938B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/176—Urban or other man-made structures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Abstract

The invention discloses a kind of method of high-resolution remote sensing image City Building function classification, its step：A, using CNN methods the building in high-resolution remote sensing image is extracted, obtain building and extract result；B, according to attribute information collating sort is carried out to POI data, and Density Estimator is carried out respectively to the POI of commerce services industry facilities land therein, public administration and public service land used and residential estate, respectively obtain the cuclear density figure of these land-use styles；C, the above-mentioned remote sensing image building information extraction result based on CNN and cuclear density figure are utilized, calculate the cuclear density average value of single building.This method is easy, it is easy to operate, efficiently solve the problem for the building functions Classification and Identification that semantic level is difficult to using remote sensing information extractive technique, function classification precision is carried out to City Building higher, the dynamic data of city function region class can be quickly and accurately provided for relevant department, be city management and the service of making rational planning for.

Description

A kind of method of high-resolution remote sensing image City Building function classification
Technical field
The present invention relates to remote sensing image classification identification technology field, a kind of high-resolution remote sensing image city is more particularly to The method of building functions classification, is particularly suitable for use in being not less than the high-resolution remote sensing image for 5m for resolution ratio.
Background technology
City Building is the important component in city, as human living with activity stable space, its transform with Renewable time affects the development in city and the change of human lives.Building can be divided into according to the use function of building The polytype such as commerce services industry facilities land, public administration and public service land used, residential land and industrial and mineral warehouse land.It is right City Building, which carries out function classification, to provide favourable foundation for urban function region division, and auxiliary government department advises to city Draw, the distribution in terms of land use, resource, population and distribution are managed and decision-making, contribute to urban sustainable development.Mesh The blowout of preceding internet data is emerged in large numbers and artificial intelligence approach is continued to develop, and has been provided for City Building function classification Data and the method support of effect, City Building function classification will turn into the trend institute studied and solve city relevant issues .
The distribution feelings of building can quickly be obtained by extracting City Building information currently with high-resolution remote sensing image Condition, helps to support city management and planning.But only be also difficult to build city by current remote sensing extractive technique of automatically classifying Build thing and carry out function classification identification, the result of extraction is often single figure spot, is lacked with semantic attribute information, it is difficult to full The demand of sufficient city planning and administration.And although the method for artificial visual interpretation can meet this demand to a certain extent, Precision is also higher, but more time-consuming.And the present invention utilizes convolutional neural networks (Convolutional Neural Networks, abbreviation CNN) method extract remote sensing image in City Building, recycle taxonomic revision POI data carry out core Density estimation (Kernel density estimation), then calculates building cuclear density average value, aids in City Building Function classification.The present invention relatively accurately realizes the extraction to remote sensing image City Building using the method for deep learning, The building functions Classification and Identification that semantic level is difficult to using remote sensing information extractive technique is also efficiently solved simultaneously Technical barrier, the dynamic data of city function region class can be quickly and accurately provided for relevant department, is city management and conjunction Reason planning service
The classification of high-resolution remote sensing image is a complicated process, using shallow structure model, such as SVMs (Support Vector Machine, abbreviation SVM) etc. classifies to it and has significant limitation, deeper structural model because Its complicated multilayered nonlinear converts and with stronger expression and modeling ability, is more suitable for the classification of high-resolution remote sensing image Extract.And although Object--oriented method is not allowed to be also easy to produce spiced salt phenomenon, but also due to experiment plot structure is complex, to segmentation The assurance of yardstick and class condition is more difficult.Deep learning is by setting up the hierarchical mode structure similar to human brain, to input data Extract step by step from bottom to high-rise feature, so as to set up the mapping relations from bottom layer signal to high-level semantic well.And Convolutional neural networks (Convolutional Neural Networks, abbreviation CNN) method in deep learning, with more multiple Miscellaneous model reduces model bias, and the degree of accuracy of statistical estimate is improved with the data of magnanimity, is declined with expansible gradient Algorithm for Solving Large-scale Optimization Problems.CNN basic structure includes two layers, and one is characterized extract layer, each neuron it is defeated Enter and be connected with the local acceptance region of preceding layer, and extract the local feature, the second is Feature Mapping layer, each calculating of network Layer is made up of multiple Feature Mappings, and each Feature Mapping is that the weights of all neurons in a plane, plane are equal.CNN its Feature monitor layer is learnt by training data, it is to avoid in explicit feature extraction, and same Feature Mapping face Neuron weights are identical, and network can be with collateral learning.Had shown that by research both domestic and external, the method for deep learning can be more to accord with The mode of conjunction human thinking completes the extraction and identification to remote sensing information, extraction result is had higher precision.Therefore, use CNN methods extract the building of high resolution image.
Data are one to point of interest (Point of Interest, abbreviation POI) can be used by people or interesting Specified point, it can be applied to multiple fields, and the interest point informations such as some hotel or gas station can be referred in ground field.Obtain The method of POI data has a lot, and relatively common is the title, classification, warp that point of interest is obtained by internet crawl technology Spend the information such as latitude, neighbouring retail shop of restaurant of hotel.To POI data carry out rationally application can Government public services, retail business, Medical services industry, manufacturing industry and it is related to the fields such as personal-location services and brings considerable value, improves ground to a certain extent Manage the integrity service level of information.
Density Estimator (Kernel density estimation) is a kind of for estimating unknown probability density function Non-parametric test method, is proposed by Rosenblatt (1955) and Emanuel Parzen (1962), also known as Parzen windows (Parzen window).Density Estimator can be used for the distribution situation for describing different shape (point, wire etc.) geometric figure, Represent density size of the geometric figure in unit area.The method of Density Estimator is compared with color density computational methods (such as sample prescription Method, based on Voronoi diagram method etc.) for, although algorithm principle is more complicated, takes more, but its numerical discretization degree is bigger, Transition of the density value between unit is smoother, the problems such as estimating in the absence of improper estimation, extreme value, is more suitable for analysis facility clothes The continuity Characteristics for spatial distribution of being engaged in.POI data is analyzed using Density Estimator, different work(more can be intuitively obtained Practicably (commerce services industry facilities land, public administration and public service land used and residential estate) is in the distribution feelings of survey region Condition.
With the development of the city with the progress of science and technology, domestic and foreign scholars to the research of city function Division increasingly Deeply, the function classification research of urban road and building in recent years is also all of concern.However, only by current remote sensing certainly Dynamic classification extractive technique is also difficult to the function classification identification that semantic level is carried out to City Building.Semantic level is carried out to city The research of classification lacks the application of remote sensing technology, general to use POI data with other statistics by being divided to survey region The method of unit grids urban function region is identified classification.In consideration of it, the present invention is extracted according to using remote sensing image building Technology, and combining geographic information Spacial Analysis technology, have invented a kind of identification of the semantic level to City Building point Class method.This method is simple and easy to apply, and the classification to destination object is more accurate.
The content of the invention
Object of the present invention is to provide a kind of method of high-resolution remote sensing image City Building function classification, side Method is easy, easy to operate, and the present invention is relatively accurately realized to remote sensing image City Building using the method for deep learning Extraction, while also efficiently solve using remote sensing information extractive technique be difficult to semantic level building functions classification The problem of identification, nicety of grading reaches 86.85%, quickly and accurately can provide the dynamic of city function region class for relevant department State data, are city management and the service of making rational planning for.
In order to realize above-mentioned purpose, the present invention uses following technical measures：
The present invention extracts the City Building in remote sensing image using CNN methods, recycles the POI data of taxonomic revision to enter Row Density Estimator, then calculates building cuclear density average value, and auxiliary carries out City Building to high-resolution remote sensing image Function classification.
A kind of method of high-resolution remote sensing image City Building function classification, its step is：
A, using CNN methods to high-definition remote sensing image data carry out building extraction, obtain the extraction knot of building Really；
Described use CNN methods carry out building extraction to high-definition remote sensing image data, obtain carrying for building Result is taken, its step is：
1) remote sensing image is pre-processed, including radiation calibration, atmospheric correction and geometric correction；
2) select the higher building sample of typical, pixel purity to be no less than 50 from above-mentioned remote sensing image, set up Building Sample Storehouse；
3) study is trained using CNN methods to the Sample Storehouse in 2), by activation functions, training contribution threshold, The parameters such as training speed, the hiding number of plies and training iterations (number of times is no less than 500 times) are configured, and set up CNN buildings Information extraction model, so as to realize high-resolution remote sensing image building to extract and obtain extracting result；
4) the too small figure spot of area in the above results is rejected, and then obtains final building and extract result.
B, according to attribute information collating sort is carried out to POI data, to commerce services industry facilities land, public administration with it is public The POI of service three kinds of function types of land used and residential estate carries out Density Estimator respectively altogether, and is layered according to cuclear density size If color, the cuclear density figure of commerce services industry facility, public administration and public service and residential function type is obtained；Its step is：
1) POI data to acquisition carries out collating sort：Crawl what is obtained from Baidu map using web crawler POI data type, its attribute information includes title, type, latitude, longitude, phone, address information, to the POI numbers of urban area According to being sorted out, commerce services industry facility, public administration and public service and residential land three types are classified as；
2) using Epanechnikov kernel functions to commerce services industry facilities land, public administration and public service land used and The POI of residential land cuclear density is calculated, and a smooth surface is covered by above each POI points, in a position Locate face value highest, as the increase face value with the distance of point is gradually reduced, the position of search radius is being equal to the distance of point The place's of putting face value is zero, using circle shaped neighborhood region, and the volume acquiescence in the space that the plane of curved surface and lower section is surrounded is equal to 1, each The density of output grid pixel is the value sum on all core surfaces for being superimposed upon grid cell, raster cell center；
Cuclear density computational methods are as follows：
If x1,x2,……xnFor independent same distribution F n observation, cuclear density is right with it from distribution density function f (x) The cumulative distribution function F (x) answered relation derivation obtains formula (1)：
Use empirical distribution functionEstimate F (x), wherein：I is indicator Function, XiDuring≤x, I (Xi≤ x)=1, I (Xi≤ x)=0, empirical distribution function Fn(x) value on one point at x, obtain with N observation x of machine variable X1,x2,……xnIn be less than or equal to x number, then divided by observation frequency n, experience is distributed letter Number is substituted into after formula (1), obtains formula (2)：
Wherein,For Density Estimator value, n is observation number, XiFor i-th of observation, K0(x) it is (non-for kernel function It is negative, integrate as 1, meet probability density property, and average is 0), h ＞ 0 is a smoothing parameter, referred to as bandwidth Or window (bandwidth).There are a variety of kernel functions, uniform, triangle, Gaussian, Epanechnikov etc..Respectively Seed nucleus function graft is see Fig. 7.
Wherein, Epanechnikov kernels are optimal under square errors sense, and loss in efficiency is minimum.Therefore, this hair Bright to use Epanechnikov to be calculated for kernel function the cuclear density of a key element, its kernel function K (x) formula is：
According to formula (3), the cuclear density of each type land used POI point datas is calculated, according to cuclear density size layer colours, Obtain the cuclear density figure of POI data commerce services industry facility, public administration and public service and residential land；
C, using remote sensing image building information extraction result and cuclear density figure based on CNN, calculate single building Cuclear density average value；Its step is：
1) vector polygon data will be converted to based on the building raster data that CNN is extracted；
2) the cuclear density figure is changed into the point for representing cuclear density size：" grid turning point " and " profit are utilized in ArcGIS Pixel value is extracted with " obtain that different size of cuclear density point can be represented；
3) count the cuclear density average value of the three class lands used inside it respectively for every solitary building.
D, cuclear density value threshold value is set to this three classes land used respectively, pass through the ratio of all kinds of cuclear density average value and the threshold value Compared with realizing the function classification of building, and then obtain high-resolution remote sensing image City Building function classification result figure.It is walked Suddenly it is：
1) commerce services industry facility, public administration and public service, the cuclear density threshold value of house are set, built for every Build, the cuclear density value of its commerce services industry facility, public administration and public service or house is more than or equal to respective threshold, point Commerce services industry facilities land, public administration and public service land used or residential land are not defined as, complete City Building Function classification；
2) the cuclear density value for two or three of type in the cuclear density value of above-mentioned three types meets threshold range Building, public administration and public service or residential land can have commerce services industry facilities land, by the classification of building Priority definition is residential land and public administration and public service land used, and specific classification standard is shown in Table 1.Cuclear density in the present invention The superposition value of all dot densities in unit is 1 longitude and latitude grid is represented, wherein decimally longitude and latitude is represented longitude and latitude.
The City Building function classification standard of table 1
By abovementioned technology, the present invention both can be more accomplished accurately to remote sensing using the method for deep learning The extraction of image City Building, efficiently solves the building that semantic level is difficult to using remote sensing information extractive technique again The problem of thing function classification identification.The invention can quickly and accurately provide the dynamic number of city function region class for relevant department According to being city management and the service of making rational planning for.
The present invention compared with prior art, with advantages below and effect：
1st, only by current remote sensing classify automatically extractive technique be also difficult to City Building carry out function classification identification, carry The result taken is often single figure spot, is lacked with semantic attribute information, it is difficult to meet the need of city planning and administration Ask.And although the method for artificial visual interpretation can meet this demand to a certain extent, precision is also higher, relatively time-consuming Arduously.The present invention solves to be difficult to the difficulty of the building functions Classification and Identification of semantic level using remote sensing information extractive technique Topic, while also having obtained the building functions classification results with compared with high-class precision.The invention more meticulously analyzes city City's information, helps precisely to divide urban function region, quickly and accurately can provide urban function region for relevant department The dynamic data of classification, is city management and the service of making rational planning for.
2nd, City Building is extracted using the CNN methods in deep learning, it is to avoid the generation of spiced salt phenomenon and The problems such as assurance to splitting yardstick and class condition is more difficult, by setting up the hierarchical mode structure similar to human brain, to input Data are extracted from bottom to high-rise feature step by step, so that the mapping relations from bottom layer signal to high-level semantic are set up well, Obtain the higher building of precision and extract result.
3rd, the present invention have chosen the Quickbird multispectral (resolution ratio reaches 2.5m) in Chaoyang District, Beijing City Asian Games Village region Exemplified by remote sensing image data, City Building is carried out to survey region using the method for the present invention and carries out function classification, effect is non- Chang Hao, classification results and artificial visual interpretation result is contrasted and precision evaluation, its total nicety of grading reaches 86.85%.
Brief description of the drawings
Fig. 1 is a kind of high-resolution remote sensing image City Building function classification method flow diagram；
Fig. 2 is the high-resolution remote sensing image schematic diagram comprising City Building employed in case study on implementation of the present invention.
The image is multispectral (resolution ratio reaches 2.5m) the remote sensing shadows of Quickbird in Chaoyang District, Beijing City Asian Games Village region As data；
Fig. 3 is the POI distribution schematic diagram used in case study on implementation of the present invention.
POI points in the figure are to crawl to obtain from Baidu map by web crawler, and its attribute information includes name The information such as title, type, latitude, longitude, phone, address, figure is the good POI data of collating sort；
Fig. 4 (a) is City Building commerce services industry facility POI cuclear density schematic diagrames.
The figure is to calculate to obtain by the POI data of commerce services industry facilities land；
Fig. 4 (b) is City Building public administration and public service POI cuclear density schematic diagrames.
The figure is to calculate to obtain by the POI data of public administration and public service land used；
Fig. 4 (c) is City Building residential land POI cuclear density schematic diagrames.
The figure is to calculate to obtain by the POI data of residential land；
Fig. 5 is that the building extracted using CNN extracts result schematic diagram.
The figure initially sets up CNN buildings extraction model and carries out building extraction, then to extracting the too small figure of area in result Spot is rejected；
Fig. 6 is the high-resolution remote sensing image City Building function classification result signal obtained using the inventive method Figure.
Mainly have commerce services industry facilities land, public administration and public service land used, residential land and it is unfiled or Other types of land used four, classification results and artificial visual interpretation result are contrasted and precision evaluation, its total nicety of grading Reach 86.85%.
Fig. 7 is the schematic diagram of several kernel functions.
Tetra- kinds of kernel function schematic diagrames of respectively uniform, triangle, Gaussian, Epanechnikov.
Embodiment
Embodiment 1：
It can be seen from Fig. 1, a kind of method of high-resolution remote sensing image City Building function classification, its step is：
A, the use CNN methods pair of the high-resolution remote sensing image A (Fig. 2) comprising City Building for giving Multispectral (resolution ratio the reaches 2.5m) remote sensing image datas of Quickbird carry out building extraction, obtain the extraction knot of building Really (such as Fig. 5).Step is as follows：
To remote sensing image pretreatment 100：Remote sensing image is pre-processed, including radiation calibration, atmospheric correction and Geometric correction etc. is handled.
101 are set up to building Sample Storehouse：The higher building of typical, pixel purity is selected from above-mentioned remote sensing image 80, sample (sample number that present case is used), sets up building Sample Storehouse.
102 are set up to CNN City Building extraction models：The Sample Storehouse set up to building Sample Storehouse in (unit 101) Study is trained using CNN methods, setting activation functions are logarithmic function, and training contribution threshold is 0.7, and training speed is 0.2, it is 2 to hide the number of plies, and training iterations is 1000 times, CNN building information extraction models is set up, so as to realize high-resolution Rate remote sensing image building extracts and obtains extracting result.
To extracting result optimizing 103：The too small figure spot of area in the above results is rejected, and then obtains final building Thing extracts result.
B, according to attribute information collating sort is carried out to POI data, to commerce services industry facilities land, public administration with it is public The POI of service three kinds of function types of land used and residential estate carries out Density Estimator respectively altogether, and is layered according to cuclear density size If color, the cuclear density figure (Fig. 4) of commerce services industry facility, public administration and public service and residential function type is obtained.Step It is as follows：
To POI data collating sort 104：The POI data that the present invention is obtained using being crawled from Baidu map is classified, POI attribute information includes the information such as title, type, latitude, longitude, phone, address.Original POI data has 23 types： Government organs, railway station subway station, bus station, bus stop, refueling station, parking lot, high-speed service area, financial service, Commercial mansion, retail trade, hotel, amusement and recreation, medical services, scientific research and education, incorporated business, park plaza, house Cell, integrated information, food and beverage sevice, automobile services, scenic spot, Telecommunications Services and public lavatory.
The type of original POI data is more, and for the ease of application, the present invention returns to POI data according to above standard Class, is specifically divided into commerce services industry facility, public administration and public service and house by POI classifications.
Wherein, commerce services industry facilities land, i.e. business, financial circles, food and drink lodging industry and other operating service industry are built Build and its corresponding affiliated facility land used.The present invention is by the refueling station in original POI data, financial service, commercial mansion, zero Sell this 11 kinds of industry, hotel, amusement and recreation, incorporated business, integrated information, food and beverage sevice, automobile services and Telecommunications Services Type definition is commerce services industry facilities land.
Public administration and public service land used, can be described as public administration and public service land used again, refer to be used for group of office The soil of body, journalism, science, education, culture and hygiene, scenic spot, communal facility etc..The present invention is by government's machine in original POI data Structure, railway station subway station, bus station, parking lot, medical services, scientific research and education, park plaza, scenic spot and public lavatory this 9 types are defined as public administration and public service land used.
Residential land, refers to for building the soil shared by resident's residential occupancy room, can be divided into town-property land used and Rural Housing Land.Because the present invention pertains only to City Building, residential quarters are defined as house use in original POI data Ground；
To Density Estimator 105：Using Epanechnikov kernel functions to commerce services industry facilities land, public administration with The POI of public service land used and residential land cuclear density is calculated, and a smooth song is covered by above each POI points Face, the face value highest at a position, as the increase face value with the distance of point is gradually reduced, with the distance of point etc. Face value is zero at the position of search radius, using circle shaped neighborhood region, the volume in the space that the plane of curved surface and lower section is surrounded Acquiescence is equal to 1, and the density of each output grid pixel is the value sum on all core surfaces for being superimposed upon grid cell, raster cell center；
Cuclear density computational methods are as follows：
If x1,x2,……xnFor independent same distribution F n observation, cuclear density is right with it from distribution density function f (x) The cumulative distribution function F (x) answered relation derivation obtains formula (1)：
Use empirical distribution functionEstimate F (x), wherein：I is indicator Function, XiDuring≤x, I (Xi≤ x)=1, I (Xi≤ x)=0, empirical distribution function Fn(x) value on one point at x, obtain with N observation x of machine variable X1,x2,……xnIn be less than or equal to x number, then divided by observation frequency n, experience is distributed letter Number is substituted into after formula (1), obtains formula (2)：
Wherein：For Density Estimator value, n is observation number, XiFor i-th of observation, K0(x) it is kernel function, h ＞ 0 is a smoothing parameter, referred to as bandwidth or window.There are an a variety of kernel functions, uniform, triangle, Gaussian, Epanechnikov etc..The figure of various kernel functions is see Fig. 7.
Wherein, Epanechnikov kernels are optimal under square errors sense, and loss in efficiency is minimum.Therefore, this hair Bright to use Epanechnikov to be calculated for kernel function the cuclear density of a key element, its kernel function K (x) formula is：
According to formula (3), the cuclear density of each type land used POI point datas is calculated, according to cuclear density size layer colours, Obtain the cuclear density figure of POI data commerce services industry facility, public administration and public service and residential land.
C, using remote sensing image building information extraction result and cuclear density figure based on CNN, calculate single building Cuclear density average value.Step is as follows：
Turn vector 106 to extracting result grid：Vector polygon will be converted to based on the building raster data that CNN is extracted Data；
To cuclear density Collapse 107：The cuclear density figure is changed into the point for representing cuclear density size：Utilized in ArcGIS " grid turning point " and " extracting pixel value using point " obtains that different size of cuclear density point can be represented；
To building cuclear density average value statistics 108：Used for three classes that every solitary building counts inside it respectively The cuclear density average value on ground.
D, cuclear density value threshold value is set to this three classes land used respectively, pass through the ratio of all kinds of cuclear density average value and the threshold value Compared with, the function classification of building is realized, and then acquisition high-resolution remote sensing image City Building function classification result figure (is schemed 6).Step is as follows：
To City Building function classification 109, complete commerce services industry facility, public administration and public service, house and use Classify on ground.It is respectively 11,000 to set commerce services industry facility, public administration and public service, the cuclear density threshold value of house, 000th, 1,800,000 and 1,300,000, for every building, its commerce services industry facility, public administration and public service or live The cuclear density value of residence is more than or equal to respective threshold, be respectively defined as commerce services industry facilities land, public administration with it is public Land used or residential land are serviced, so as to realize the function classification of City Building；
110 are set up to City Building function classification standard.For two kinds in the cuclear density value of above-mentioned three types or The cuclear density value of three types meets the building of threshold range, and public administration can have business with public service or residential land Service trade facilities land, is residential land and public administration and public service land used by the classification priority definition of building, specifically Criteria for classification is shown in Table 1.Cuclear density in the present invention represents the superposition value of all dot densities in unit is 1 longitude and latitude grid, Wherein decimally longitude and latitude is represented longitude and latitude.
The City Building function classification standard of table 1
Function classification is carried out to City Building using the method introduced of the present invention, and by classification results and artificial visual solution Translate result to be contrasted and precision evaluation, its total nicety of grading reaches 86.85%.
Remote sensing technology, internet POI data and Spatial Data Analysis are effectively bonded together by the present invention, more carefully Analyze urban information with causing, realize the classification of City Building function, improve the accuracy rate and confidence level of Classification and Identification, Help precisely to divide urban function region.In research from now on, city is realized using the method system of the present invention The Classification and Identification of building different type land used.
By above-mentioned technical measures, the present invention is relatively accurately realized to remote sensing image using the method for deep learning The extraction of City Building, while also efficiently solving the building that semantic level is difficult to using remote sensing information extractive technique The problem of thing function classification identification, nicety of grading reaches 86.85%, quickly and accurately can provide city function for relevant department The dynamic data of region class, is city management and the service of making rational planning for.
It should be pointed out that embodiment described above can make those skilled in the art that this hair is more fully understood It is bright, but do not limit the invention in any way.Therefore, it will be appreciated by those skilled in the art that still can be to present invention progress Modification or equivalent substitution；And technical scheme and its improvement of all spirit and technical spirit that do not depart from the present invention, it all should Cover among the protection domain of patent of the present invention.

Claims (1)

1. a kind of method of high-resolution remote sensing image City Building function classification, its step is：
A, using convolutional neural networks method to high-definition remote sensing image data carry out building extraction, obtain carrying for building Take result；Its step is：
1）Remote sensing image is pre-processed, including radiation calibration, atmospheric correction and geometric correction；
2）Select the building sample of typical, pixel purity to be no less than 50 from above-mentioned remote sensing image, set up building sample This storehouse；
3）To step 2）In Sample Storehouse be trained study using CNN methods, by activation functions, training contribution threshold, Training speed, the hiding number of plies and training iterations, number of times are no less than 500 times, and parameter is configured, and set up CNN buildings letter Extraction model is ceased, realizes that high-resolution remote sensing image building is extracted and obtains extracting result；
4）The too small figure spot of area in the above results is rejected, building is obtained and extracts result；
B, according to attribute information collating sort is carried out to POI data, to commerce services industry facilities land, public administration and public clothes The POI of business three kinds of function types of land used and residential estate carries out Density Estimator respectively, obtains commerce services industry facility, public pipe Reason and public service and the cuclear density figure of residential function type；Its step is：
1）POI data to acquisition carries out collating sort：The POI numbers obtained are crawled from Baidu map using web crawler According to type, its attribute information includes title, type, latitude, longitude, phone, address information, and the POI data to urban area is carried out Sort out, be classified as commerce services industry facility, public administration and public service and residential land three types；
2）Using Epanechnikov kernel functions to commerce services industry facilities land, public administration and public service land used and house The POI of land used cuclear density is calculated, and a smooth surface, the table at a position are covered by above each POI points Face amount highest, as the increase face value with the distance of point is gradually reduced, is being equal at the position of search radius with the distance of point Face value is zero, using circle shaped neighborhood region, and the volume acquiescence in the space that the plane of curved surface and lower section is surrounded is equal to 1, each output The density of grid cell, raster cell is the value sum on all core surfaces for being superimposed upon grid cell, raster cell center；
Cuclear density computational methods are as follows：
If x1,x2,……xnFor independent same distribution F'snIndividual observation, cuclear density is from distribution density functionf(x)Corresponding Cumulative distribution functionF(x)Relation derivation obtain formula（1）：
（1）
Use empirical distribution functionEstimationF(x), wherein：IIt is indicator function,When,=1,=0, empirical distribution functionF n (x)On one pointxThe value at place, obtains stochastic variableX'sn Individual observation x1,x2,……xnIn be less than or equal toxNumber, then divided by observation frequencyn, empirical distribution function is substituted into formula (1) after, formula is obtained（2）：
（2）
Wherein：For Density Estimator value,nFor observation number,X i ForiIndividual observation,K 0 (x)For kernel function,h＞ 0 is One smoothing parameter, referred to as bandwidth or window；
Wherein：Epanechnikov is used to be calculated for kernel function the cuclear density of a key element, its kernel functionK(x)Formula is：
(3)
According to formula（3）, the cuclear density of each type land used POI point datas is calculated, according to cuclear density size layer colours, is obtained The cuclear density figure of POI data commerce services industry facility, public administration and public service and residential land；
C, using remote sensing image building information extraction result and cuclear density figure based on CNN, the core for calculating single building is close Spend average value；Its step is：
1）Vector polygon data will be converted to based on the building raster data that CNN is extracted；
2）The cuclear density figure is changed into the point for representing cuclear density size：Extracted in ArcGIS using grid turning point and using point Pixel, which is worth to, can represent different size of cuclear density point；
3）Count the cuclear density average value of the three class lands used inside it respectively for every solitary building；
D, respectively to this three classes land used set cuclear density value threshold value, by the comparison of all kinds of cuclear density average value Yu the threshold value, The function classification of building is realized, and then obtains high-resolution remote sensing image City Building function classification result figure, its step It is：
1）Commerce services industry facility, public administration and public service, the cuclear density threshold value of house are set, built for every, its The cuclear density value of commerce services industry facility, public administration and public service or house is more than or equal to respective threshold, fixed respectively Justice is commerce services industry facilities land, public administration and public service land used or residential land, completes the function of City Building Classification；
2）Cuclear density value for two or three of type in the cuclear density value of above-mentioned three types meets building for threshold range Thing is built, public administration can have commerce services industry facilities land with public service or residential land, and the classification of building is preferential It is defined as residential land and public administration and public service land used.
CN201710425766.8A 2017-06-08 2017-06-08 high-resolution remote sensing image urban building function classification method Active CN107247938B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710425766.8A CN107247938B (en) 2017-06-08 2017-06-08 high-resolution remote sensing image urban building function classification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710425766.8A CN107247938B (en) 2017-06-08 2017-06-08 high-resolution remote sensing image urban building function classification method

Publications (2)

Publication Number Publication Date
CN107247938A true CN107247938A (en) 2017-10-13
CN107247938B CN107247938B (en) 2019-12-06

Family

ID=60019057

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710425766.8A Active CN107247938B (en) 2017-06-08 2017-06-08 high-resolution remote sensing image urban building function classification method

Country Status (1)

Country Link
CN (1) CN107247938B (en)

Cited By (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108052876A (en) * 2017-11-28 2018-05-18 广东数相智能科技有限公司 Regional development appraisal procedure and device based on image identification
CN108182253A (en) * 2017-12-29 2018-06-19 百度在线网络技术（北京）有限公司 For generating the method and apparatus of information
CN108764193A (en) * 2018-06-04 2018-11-06 北京师范大学 Merge the city function limited region dividing method of POI and remote sensing image
CN109086715A (en) * 2018-07-31 2018-12-25 天图软件科技有限公司 The method of system, device and data processing based on the distant data middleware of admittance
CN109189917A (en) * 2018-06-27 2019-01-11 华南师范大学 A kind of city function limited region dividing method and system merging landscape and social characteristic
CN109325085A (en) * 2018-08-08 2019-02-12 中南大学 A kind of urban land identification of function and change detecting method
CN109446992A (en) * 2018-10-30 2019-03-08 苏州中科天启遥感科技有限公司 Remote sensing image building extracting method and system, storage medium, electronic equipment based on deep learning
CN109493119A (en) * 2018-10-19 2019-03-19 南京图申图信息科技有限公司 A kind of city commercial center identification method and system based on POI data
CN109993753A (en) * 2019-03-15 2019-07-09 北京大学 The dividing method and device of urban function region in remote sensing image
CN110147779A (en) * 2019-05-27 2019-08-20 宁夏云图勘测规划有限公司 A kind of method of remote sensing image City Building function classification
CN110298211A (en) * 2018-03-21 2019-10-01 北京大学 A kind of Methods Deriving Drainage Network based on deep learning and high-resolution remote sensing image
CN110472559A (en) * 2019-08-13 2019-11-19 苏州中科天启遥感科技有限公司 A kind of remote sensing image construction area land use attribute space moving method
CN110555352A (en) * 2018-06-04 2019-12-10 百度在线网络技术（北京）有限公司 interest point identification method, device, server and storage medium
CN110580443A (en) * 2019-06-19 2019-12-17 深圳大学 Low-altitude near-real-time building earthquake damage assessment method
CN110633890A (en) * 2019-08-06 2019-12-31 广东晟腾地信科技有限公司 Land utilization efficiency judgment method and system
CN110866156A (en) * 2019-11-26 2020-03-06 北京明略软件系统有限公司 Method, device, equipment and medium for identifying functional park based on social data
CN111353665A (en) * 2018-12-21 2020-06-30 北京航天泰坦科技股份有限公司 Method for evaluating city expansion space-time evolution characteristics based on remote sensing data
CN111382330A (en) * 2020-03-10 2020-07-07 智慧足迹数据科技有限公司 Land property identification method and device, electronic equipment and storage medium
CN111563448A (en) * 2020-04-30 2020-08-21 北京百度网讯科技有限公司 Method and device for detecting illegal building, electronic equipment and storage medium
CN111582146A (en) * 2020-05-06 2020-08-25 宁波大学 High-resolution remote sensing image city function partitioning method based on multi-feature fusion
CN111582194A (en) * 2020-05-12 2020-08-25 吉林大学 Multi-temporal high-resolution remote sensing image building extraction method based on multi-feature LSTM network
CN111797188A (en) * 2020-06-28 2020-10-20 武汉大学 Urban functional area quantitative identification method based on open-source geographic space vector data
CN111832527A (en) * 2020-07-23 2020-10-27 中国科学院空天信息创新研究院 Residential community extraction and type identification method based on remote sensing and social perception data
CN112015841A (en) * 2020-08-28 2020-12-01 南昌市国土资源勘测规划院有限公司 Space planning cardinality conversion method based on big data and image recognition
CN112132729A (en) * 2020-10-19 2020-12-25 林文旭 Intelligent city planning system
CN112925764A (en) * 2019-12-10 2021-06-08 中国再保险(集团)股份有限公司 Industrial risk exposure database of disaster giant model and construction method thereof
CN113033484A (en) * 2021-04-21 2021-06-25 河北工程大学 Urban classification method for unmanned aerial vehicle emergency network deployment
CN113505842A (en) * 2021-07-21 2021-10-15 中国科学院空天信息创新研究院 Automatic extraction method suitable for large-scale regional remote sensing image urban building
CN113672788A (en) * 2021-07-22 2021-11-19 东南大学 Urban building function classification method based on multi-source data and weight coefficient method
CN114694034A (en) * 2020-12-28 2022-07-01 株式会社协同创新技术 Method and apparatus for providing educational service using artificial intelligence-based satellite image
CN116543310A (en) * 2023-06-30 2023-08-04 眉山环天智慧科技有限公司 Road line extraction method based on Voronoi diagram and kernel density

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101126812A (en) * 2007-09-27 2008-02-20 武汉大学 High resolution ratio remote-sensing image division and classification and variety detection integration method
CN101510311A (en) * 2009-03-05 2009-08-19 浙江大学 Method for rapidly sorting a large amount of building side elevation images based on GPS information
CN102496034A (en) * 2011-11-29 2012-06-13 南京师范大学 High-spatial resolution remote-sensing image bag-of-word classification method based on linear words
CN103364781A (en) * 2012-04-11 2013-10-23 南京财经大学 Remote sensing data and geographical information system-based grainfield ground reference point screening method
CN104103076A (en) * 2014-07-29 2014-10-15 环境保护部卫星环境应用中心 Nuclear power plant planned restricted zone remote sensing inspecting method based on high-resolution remote sensing images
CN104200223A (en) * 2014-08-29 2014-12-10 中国科学院城市环境研究所 Urban function area identifying method based on high-resolution remote sensing image
CN105138668A (en) * 2015-09-06 2015-12-09 中山大学 Urban business center and retailing format concentrated area identification method based on POI data
CN105184308A (en) * 2015-08-03 2015-12-23 北京航空航天大学 Remote sensing image building detection and classification method based on global optimization decision
CN106021499A (en) * 2016-05-20 2016-10-12 中国农业大学 Construction land classification method and device based on geographic information of volunteer
CN106529600A (en) * 2016-11-16 2017-03-22 桂林理工大学 SVM-based recognition method of building angular points in high-resolution optical image

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101126812A (en) * 2007-09-27 2008-02-20 武汉大学 High resolution ratio remote-sensing image division and classification and variety detection integration method
CN101510311A (en) * 2009-03-05 2009-08-19 浙江大学 Method for rapidly sorting a large amount of building side elevation images based on GPS information
CN102496034A (en) * 2011-11-29 2012-06-13 南京师范大学 High-spatial resolution remote-sensing image bag-of-word classification method based on linear words
CN103364781A (en) * 2012-04-11 2013-10-23 南京财经大学 Remote sensing data and geographical information system-based grainfield ground reference point screening method
CN104103076A (en) * 2014-07-29 2014-10-15 环境保护部卫星环境应用中心 Nuclear power plant planned restricted zone remote sensing inspecting method based on high-resolution remote sensing images
CN104200223A (en) * 2014-08-29 2014-12-10 中国科学院城市环境研究所 Urban function area identifying method based on high-resolution remote sensing image
CN105184308A (en) * 2015-08-03 2015-12-23 北京航空航天大学 Remote sensing image building detection and classification method based on global optimization decision
CN105138668A (en) * 2015-09-06 2015-12-09 中山大学 Urban business center and retailing format concentrated area identification method based on POI data
CN106021499A (en) * 2016-05-20 2016-10-12 中国农业大学 Construction land classification method and device based on geographic information of volunteer
CN106529600A (en) * 2016-11-16 2017-03-22 桂林理工大学 SVM-based recognition method of building angular points in high-resolution optical image

Cited By (49)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108052876B (en) * 2017-11-28 2022-02-11 广东数相智能科技有限公司 Regional development assessment method and device based on image recognition
CN108052876A (en) * 2017-11-28 2018-05-18 广东数相智能科技有限公司 Regional development appraisal procedure and device based on image identification
CN108182253A (en) * 2017-12-29 2018-06-19 百度在线网络技术（北京）有限公司 For generating the method and apparatus of information
CN108182253B (en) * 2017-12-29 2021-12-28 百度在线网络技术（北京）有限公司 Method and apparatus for generating information
CN110298211A (en) * 2018-03-21 2019-10-01 北京大学 A kind of Methods Deriving Drainage Network based on deep learning and high-resolution remote sensing image
CN108764193B (en) * 2018-06-04 2022-02-01 北京师范大学 Urban functional area dividing method fusing POI and remote sensing image
CN108764193A (en) * 2018-06-04 2018-11-06 北京师范大学 Merge the city function limited region dividing method of POI and remote sensing image
CN110555352A (en) * 2018-06-04 2019-12-10 百度在线网络技术（北京）有限公司 interest point identification method, device, server and storage medium
CN109189917B (en) * 2018-06-27 2020-07-28 华南师范大学 City functional zone division method and system integrating landscape and social characteristics
CN109189917A (en) * 2018-06-27 2019-01-11 华南师范大学 A kind of city function limited region dividing method and system merging landscape and social characteristic
CN109086715A (en) * 2018-07-31 2018-12-25 天图软件科技有限公司 The method of system, device and data processing based on the distant data middleware of admittance
CN109325085A (en) * 2018-08-08 2019-02-12 中南大学 A kind of urban land identification of function and change detecting method
CN109493119A (en) * 2018-10-19 2019-03-19 南京图申图信息科技有限公司 A kind of city commercial center identification method and system based on POI data
CN109493119B (en) * 2018-10-19 2020-06-23 南京图申图信息科技有限公司 POI data-based urban business center identification method and system
CN109446992B (en) * 2018-10-30 2022-06-17 苏州中科天启遥感科技有限公司 Remote sensing image building extraction method and system based on deep learning, storage medium and electronic equipment
CN109446992A (en) * 2018-10-30 2019-03-08 苏州中科天启遥感科技有限公司 Remote sensing image building extracting method and system, storage medium, electronic equipment based on deep learning
CN111353665A (en) * 2018-12-21 2020-06-30 北京航天泰坦科技股份有限公司 Method for evaluating city expansion space-time evolution characteristics based on remote sensing data
CN109993753B (en) * 2019-03-15 2021-03-23 北京大学 Method and device for segmenting urban functional area in remote sensing image
CN109993753A (en) * 2019-03-15 2019-07-09 北京大学 The dividing method and device of urban function region in remote sensing image
CN110147779A (en) * 2019-05-27 2019-08-20 宁夏云图勘测规划有限公司 A kind of method of remote sensing image City Building function classification
CN110580443B (en) * 2019-06-19 2022-03-08 深圳大学 Low-altitude near-real-time building earthquake damage assessment method
CN110580443A (en) * 2019-06-19 2019-12-17 深圳大学 Low-altitude near-real-time building earthquake damage assessment method
CN110633890A (en) * 2019-08-06 2019-12-31 广东晟腾地信科技有限公司 Land utilization efficiency judgment method and system
CN110472559B (en) * 2019-08-13 2023-06-23 苏州中科天启遥感科技有限公司 Remote sensing image building area land utilization attribute space migration method
CN110472559A (en) * 2019-08-13 2019-11-19 苏州中科天启遥感科技有限公司 A kind of remote sensing image construction area land use attribute space moving method
CN110866156B (en) * 2019-11-26 2022-05-17 北京明略软件系统有限公司 Method, device, equipment and medium for identifying functional park based on social data
CN110866156A (en) * 2019-11-26 2020-03-06 北京明略软件系统有限公司 Method, device, equipment and medium for identifying functional park based on social data
CN112925764B (en) * 2019-12-10 2023-09-12 中国再保险(集团)股份有限公司 Industry risk exposure database of disaster model and construction method thereof
CN112925764A (en) * 2019-12-10 2021-06-08 中国再保险(集团)股份有限公司 Industrial risk exposure database of disaster giant model and construction method thereof
CN111382330A (en) * 2020-03-10 2020-07-07 智慧足迹数据科技有限公司 Land property identification method and device, electronic equipment and storage medium
CN111563448A (en) * 2020-04-30 2020-08-21 北京百度网讯科技有限公司 Method and device for detecting illegal building, electronic equipment and storage medium
CN111582146B (en) * 2020-05-06 2021-04-13 宁波大学 High-resolution remote sensing image city function partitioning method based on multi-feature fusion
CN111582146A (en) * 2020-05-06 2020-08-25 宁波大学 High-resolution remote sensing image city function partitioning method based on multi-feature fusion
CN111582194A (en) * 2020-05-12 2020-08-25 吉林大学 Multi-temporal high-resolution remote sensing image building extraction method based on multi-feature LSTM network
CN111582194B (en) * 2020-05-12 2022-03-29 吉林大学 Multi-temporal high-resolution remote sensing image building extraction method based on multi-feature LSTM network
CN111797188A (en) * 2020-06-28 2020-10-20 武汉大学 Urban functional area quantitative identification method based on open-source geographic space vector data
CN111797188B (en) * 2020-06-28 2024-03-01 武汉大学 Urban functional area quantitative identification method based on open source geospatial vector data
CN111832527B (en) * 2020-07-23 2023-09-05 中国科学院空天信息创新研究院 Residential area extraction and type identification method based on remote sensing and social sensing data
CN111832527A (en) * 2020-07-23 2020-10-27 中国科学院空天信息创新研究院 Residential community extraction and type identification method based on remote sensing and social perception data
CN112015841A (en) * 2020-08-28 2020-12-01 南昌市国土资源勘测规划院有限公司 Space planning cardinality conversion method based on big data and image recognition
CN112132729A (en) * 2020-10-19 2020-12-25 林文旭 Intelligent city planning system
CN114694034A (en) * 2020-12-28 2022-07-01 株式会社协同创新技术 Method and apparatus for providing educational service using artificial intelligence-based satellite image
CN113033484A (en) * 2021-04-21 2021-06-25 河北工程大学 Urban classification method for unmanned aerial vehicle emergency network deployment
CN113505842B (en) * 2021-07-21 2024-01-05 中国科学院空天信息创新研究院 Automatic urban building extraction method suitable for large-scale regional remote sensing image
CN113505842A (en) * 2021-07-21 2021-10-15 中国科学院空天信息创新研究院 Automatic extraction method suitable for large-scale regional remote sensing image urban building
CN113672788A (en) * 2021-07-22 2021-11-19 东南大学 Urban building function classification method based on multi-source data and weight coefficient method
CN113672788B (en) * 2021-07-22 2024-04-09 东南大学 Urban building function classification method based on multi-source data and weight coefficient method
CN116543310A (en) * 2023-06-30 2023-08-04 眉山环天智慧科技有限公司 Road line extraction method based on Voronoi diagram and kernel density
CN116543310B (en) * 2023-06-30 2023-10-31 眉山环天智慧科技有限公司 Road line extraction method based on Voronoi diagram and kernel density

Also Published As

Publication number Publication date
CN107247938B (en) 2019-12-06

Similar Documents

Publication Publication Date Title
CN107247938A (en) A kind of method of high-resolution remote sensing image City Building function classification
CN109145939B (en) Semantic segmentation method for small-target sensitive dual-channel convolutional neural network
CN109508360B (en) Geographical multivariate stream data space-time autocorrelation analysis method based on cellular automaton
CN108920481B (en) Road network reconstruction method and system based on mobile phone positioning data
CN103955702B (en) SAR image terrain classification method based on depth RBF network
Grekousis et al. Modeling urban evolution using neural networks, fuzzy logic and GIS: The case of the Athens metropolitan area
CN100557626C (en) Image partition method based on immune spectrum clustering
CN109272170B (en) A kind of traffic zone dividing system based on Louvain algorithm
CN110084294A (en) A kind of Remote Image Classification based on multiple dimensioned depth characteristic
CN108052966A (en) Remote sensing images scene based on convolutional neural networks automatically extracts and sorting technique
CN103888541B (en) Method and system for discovering cells fused with topology potential and spectral clustering
CN107480620A (en) Remote sensing images automatic target recognition method based on heterogeneous characteristic fusion
CN108108808A (en) A kind of position predicting method and device based on depth belief network
CN112950780B (en) Intelligent network map generation method and system based on remote sensing image
CN106683102A (en) SAR image segmentation method based on ridgelet filters and convolution structure model
CN112347970A (en) Remote sensing image ground object identification method based on graph convolution neural network
CN102902976A (en) Image scene classification method based on target and space relationship characteristics
CN106056609A (en) Method based on DBNMI model for realizing automatic annotation of remote sensing image
CN109426813B (en) Remote sensing image user-defined interest point extraction method based on fuzzy clustering and neural network model
Zhuang et al. Integrating a deep forest algorithm with vector‐based cellular automata for urban land change simulation
CN104574368B (en) A kind of adaptive kernel clustering image partition method
CN104331711B (en) SAR image recognition methods based on multiple dimensioned fuzzy mearue and semi-supervised learning
Ozyirmidokuz et al. A data mining based approach to a firm's marketing channel
CN107451617A (en) One kind figure transduction semisupervised classification method
CN116662468A (en) Urban functional area identification method and system based on geographic object space mode characteristics

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant